A Monte Carlo simulation of the flow network reliability using importance and stratified sampling

We consider the evaluation of the flow network reliability parameter. Because its exact computation has exponential time complexity, simulation methods are alternatives used to evaluate large networks. In this paper, we use the state space decomposition methodology of Doulliez and Jamoulle in order to construct a new simulation method which combines the importance and the stratified Monte Carlo principles. We show that the related estimator belongs to the variance reduction family and it offers more accurate estimates than those obtained by a previous stratified sampling technique based on the same decomposition, By experimental results, we show the interest of the proposed method when compared to previous methods.